1 / 28

An Efficient Branch-and-Bound Algorithm for Optimal Human Pose Estimation

An Efficient Branch-and-Bound Algorithm for Optimal Human Pose Estimation. OUTLINE. Introduction The Human Pose Estimation MRF Branch and bound Experiments Conclusion. Introduction. Introduction. Simple representations(like tree or star models) can be efficiently applied

bruno
Download Presentation

An Efficient Branch-and-Bound Algorithm for Optimal Human Pose Estimation

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. An Efficient Branch-and-Bound Algorithm for Optimal Human Pose Estimation

  2. OUTLINE • Introduction • The Human Pose Estimation • MRF • Branch and bound • Experiments • Conclusion

  3. Introduction

  4. Introduction • Simple representations(like tree or star models) can be efficiently applied • Misclassification errors • Models with rich representations (i.e., loopy graphical models) are theoretically more robust • Time complexity may increase dramatically

  5. Introduction • Propose an efficient and exact inference algorithm • Based on branch-and-bound • Solving MAP inference on general MRF • Contribution • Relaxing the loopy model into a mixture of star-models • A special data structure (BMT) and an efficient search routine(OBMS)

  6. Introduction

  7. OUTLINE • Introduction • The Human Pose Estimation • MRF • Branch and bound • Experiments • Conclusion

  8. MRF in loopy model • Equivalent to the Maximum a Posteriori(MAP) inference problem • finds the best assignment • 1

  9. MRF in mix_star model

  10. MRF in mix_star model

  11. MRF in mix_star model

  12. OUTLINE • Introduction • The Human Pose Estimation • MRF • Branch and bound • Experiments • Conclusion

  13. Branch and bound • The MAP inference problem is hard • The hypothesis space is large • DP work well on tree models • Can’t be applied due to the complicated pair-wise relationships • Use the likelihood do branch into two subspace • The upper bound of the value of the MAP assignment is used as “the likelihood”

  14. BMT

  15. Branch and bound

  16. Branch and bound

  17. Branch and bound

  18. Branch and bound

  19. Branch and bound

  20. OUTLINE • Introduction • The Human Pose Estimation • MRF • Branch and bound • Experiments • Conclusion

  21. Experiments • This analysis using the Stretchable Models (SM) • Device • 64-bit 16-Core Intel(R) • Xeon(R) 2.40GHz CPU • 48GB RAM • Algorithm • C++ • Matlab

  22. Experiments

  23. Experiments • Buffy • X: Our method • Y: CP method • Green: Without OBMS • Red: Full BB approach

  24. OUTLINE • Introduction • The Human Pose Estimation • MRF • Branch and bound • Experiments • Conclusion

  25. Conclusion • Shown that our efficient and exact inference algorithm • An interesting future research direction • Learning parameters of complex models to achieve accurate performance • Source Codes

  26. Thank you

  27. http://www.youtube.com/watch?v=cAAhHSaGgXA

More Related